Received: 28 March 2019 | Revised: 2 July 2019 | Accepted: 3 July 2019 DOI: 10.1111/jvs.12796

RESEARCH ARTICLE Journal of Vegetation Science

Fire severity and changing composition of forest understory communities

Jens T. Stevens1 | Jesse E. D. Miller2 | Paula J. Fornwalt3

1U.S. Geological Survey, Landscapes Field Station, Santa Fe, NM, Abstract USA Questions: Gradients of fire severity in dry conifer forests can be associated with 2 Department of Biology, Stanford University, variation in understory floristic composition. Recent work in dry conifer forests in Stanford, CA, USA 3USDA Forest Service, Rocky Mountain , USA, has suggested that more severely burned stands contain more ther‐ Research Station, Fort Collins, CO, USA mophilic taxa (those associated with warmer and drier conditions), and that forest

Correspondence disturbance may therefore accelerate floristic shifts already underway due to climate Jens T. Stevens, U.S. Geological Survey, New change. However, it remains unknown how rapidly thermophilic taxa shifts occur Mexico Landscapes Field Station, Santa Fe, NM 87508, USA. following disturbance, how long such shifts are likely to persist, and how different Email: [email protected] thermophilic post‐disturbance communities are from pre‐disturbance communities.

Funding information Location: Front Range, USA. This paper was written and prepared using Methods: We investigated these questions using a unique 15‐year vegetation plot U.S. Government funds and as such it is in the public domain and not subject to dataset that captures pre‐ and post‐fire understory community composition across copyright. a gradient of fire severity in dry conifer forests, classifying taxa using the biogeo‐ Co‐ordinating Editor: John Morgan graphic affinity concept. Results: Thermophilization (defined here as a decrease in the ratio of cool‐mesic taxa to warm‐xeric taxa, based on biogeographic affinity of paleobotanical lineages) was observed as early as one year post‐fire for all fire severity classes, but was stronger at sites that burned at higher severity. The ratio of cool‐mesic to warm‐xeric taxa recov‐ ered to pre‐fire levels within 10 years in stands that burned at low severity, but not in stands that burned at moderate or high severity. The process of thermophilization after high‐severity fire appears to be driven primarily by the gain of warm‐xeric taxa that were absent before the fire, but losses of cool‐mesic taxa, which did not return during the duration of the study, also played a role. Conclusions: Decreases in canopy cover appear to be a main contributor to under‐ story thermophilization. Fine‐scale heterogeneity in post‐fire forest structure is likely an important driver of floristic diversity, creating the microclimatic variation neces‐ sary to maintain floristic refugia for mal‐adapted to increasingly warm and dry conditions.

KEYWORDS biodiversity, biogeographic affinity, Colorado, dry conifer forests, fire, Hayman Fire, thermophilization, understory

J Veg Sci. 2019;00:1–11. wileyonlinelibrary.com/journal/jvs © 2019 International Association | 1 for Vegetation Science 2 STEVENS et al. | Journal of Vegetation Science

1 | INTRODUCTION species could represent biodiversity losses of substantial conserva‐ tion concern. Generalizable patterns of plant community responses to disturbance Ongoing increases in fire severity, and particularly the increasing can be difficult to discern, in part because plant communities are incidence of uncharacteristically large, spatially contiguous patches composed of many species that have diverse adaptations to distur‐ of high‐severity fire in dry conifer forests of some regions (Singleton, bance (Grime, 1977). Species niches are often conserved through Thode, Sánchez Meador, & Iniguez, 2019; Stevens et al., 2017), have evolutionary time (Ackerly, 2003), and environmental conditions raised concern that forest stands burning at high severity over large present at the time of a particular clade's diversification are likely areas may lack the resilience to return to a forested condition, be‐ related to the contemporary niche space where those clades occur cause the dominant species require seeds produced by living trees (Ackerly, 2009). Therefore, lineages of similar biogeographic affin‐ in order to regenerate (Johnstone et al., 2016). The time required ities (appearances in the paleobotanical record at a similar place for mature trees to reestablish and create the heterogeneous un‐ and time) may be more likely to have relatively similar response to derstory microclimate likely associated with historical forest condi‐ environmental conditions, including those induced by disturbance tions (Dodson & Peterson, 2010) is unknown in many forest types. (Harrison & Grace, 2007). Longitudinal studies of post‐fire understory communities are rare, In forests, canopy cover is a primary driver of understory species but when present they offer the opportunity to document the resil‐ composition (Hart & Chen, 2006). Canopy cover mediates the un‐ ience of understory communities to different degrees of disturbance derstory microclimate; for instance, canopy disturbance can cause severity over time. increases in understory temperature, wind speed and transpirational In this paper we apply the biogeographic affinity concept to doc‐ demand (Ma, Concilio, Oakley, North, & Chen, 2010). Alterations ument thermophilization within a unique longitudinal study of vas‐ to canopy cover and the understory microclimate can in turn cre‐ cular understory vegetation spanning 15 years around the Hayman ate conditions that favor certain understory plant species, includ‐ Fire, a large fire that occurred in dry conifer forests of the Colorado ing species that can tolerate the increased temperatures and water Rocky Mountains, USA (Abella & Fornwalt, 2015; Fornwalt, stress that sometimes accompany canopy removal (Ma et al., 2010). Kaufmann, & Stohlgren, 2010). While previous research in the Sierra Particularly in climatically drier forests, warmer and more water‐ Nevada, California, USA (Stevens et al., 2015) has documented stressed microclimates in the post‐disturbance landscape can cause stronger thermophilization following high‐severity fire compared increases in species adapted to warmer, more xeric conditions and with low‐severity fire or forest thinning, this process has not been decreases in species with affinities for cooler, more mesic conditions, examined for longer than three years post‐fire, nor has the relative a phenomenon that has been termed “thermophilization” (Gottfried contribution of extinction and colonization to the thermophilization et al., 2012). In general, species that decline after disturbance tend process been examined, due to the rarity of pre‐disturbance data. to have biogeographic affinities for higher latitudes, while species We examine temporal trends in biogeographic affinity of the under‐ that increase after disturbance tend to be species with more equato‐ story vegetation community on the Hayman Fire to ask (a) whether rial biogeographic affinities (De Frenne et al., 2013; Stevens, Safford, the short‐term thermophilization caused by high‐severity fire occurs Harrison, & Latimer, 2015). in dry conifer forests of the Rocky Mountains to a similar extent as Forest canopy disturbance, particularly from high‐severity fire in the Sierra Nevada; (b) how long understory thermophilization per‐ that removes an entire tree canopy within a given area, can lead sists following disturbance of different severities; and (c) whether to dramatic and rapid understory plant community shifts, poten‐ thermophilization is driven primarily by the extinction of cool‐mesic tially accelerating the shifts in vegetation expected with climate affinity taxa, the colonization of warm‐xeric affinity taxa, or both. By warming (Miller & Safford, 2019; Stevens et al., 2015). This rapid investigating these previously unaddressed questions, we hope to thermophilization via disturbance is distinct from gradual climate increase our understanding of the resilience of understory vegeta‐ warming and forest cover reductions, which can lead to varying de‐ tion to gradients in disturbance severity over time. grees of protracted understory plant community thermophilization (De Frenne et al., 2013; Gottfried et al., 2012). And while general 2 | METHODS changes in understory community composition associated with fire have been observed for at least 20–30 years post‐fire in conifer for‐ 2.1 | Study site ests (Coop, Massatti, & Schoettle, 2010; Romme, Whitby, Tinker, & Turner, 2016; Webster & Halpern, 2010), it remains unknown how Our study was conducted at a 400‐ha site in the Pike National Forest long post‐fire thermophilization might persist, although post‐fire (39.14° N, 105.24° W), approximately 60 km southwest of Denver, successional trajectories likely play a role (Stevens, Collins, Miller, CO, USA (Figure 1). Elevations at the site range from about 2,300 to North, & Stephens, 2017). Further, although fire has been shown to 2,500 m. The climate is warm and dry — over the 1996–2012 study cause shifts towards plant communities with a greater proportion of period, average January temperature was about −4°C, average July warm‐xeric species, it is unknown whether this pattern is driven by temperature was about 17°C, and average annual precipitation was gains of such species, losses of cool‐mesic species, or both (Stevens about 320 mm (Appendix S1; Asherin, 2016). Soils, derived from et al., 2015). Long‐term landscape‐scale extirpations of cool‐mesic Pikes Peak granite, are poorly developed. Historical fire intervals STEVENS et al. 3 Journal of Vegetation Science |

FIGURE 1 Location of the 400‐ha study site within the 2002 Hayman Fire, Colorado, USA (left). Location and severity of the 20 upland 0.1‐ha plots within the study site (right). Post‐fire aerial imagery is shown in the background (imagery sources: Esri, DigitalGlobe, GeoEye, Earthstar Geographics, CNES/Airbus DS, USDA, USGS, AeroGRID, IGN, and the GIS User Community). This figure was adapted from Fornwalt et al. (2018) for individual –Pseudotsuga menziesii forest stands at years in eventual fire severity (χ2 = 1.27, df = 2, p = 0.53) or in the pre‐ and surrounding the site varied from very short (≤10 years) to very fire ratio of our two floristic groups described below (t = 0, df = 18, long (>100 years) (Brown, Kaufmann, Shepperd, 1999). Historical p = 0.998), so for simplicity hereafter we ascribe all pre‐fire sampling fires likely included patches of high‐severity fire effects within a to 1997. Composition was assessed by recording the identities of all matrix of lower‐severity fire effects (Brown et al. 1999), with high‐ taxa in the plots. All graminoids, forbs, and were included in severity patches likely not exceeding about 100 ha (Romme, Veblen, our surveys, but trees were not. Unknown taxa were collected for Kaufmann, Sherriff, & Regan, 2003). This site experienced few wild‐ later identification. While most taxa were ultimately identified to the fires beginning in the late 1800s, coincident with Euro‐American species level (varieties and subspecies were not distinguished), some settlement and associated land‐use activities including fire suppres‐ taxa were only identified to the genus level because key morpho‐ sion, livestock grazing, and logging. logical characteristics were not sufficiently developed. Carex spp., In 2002, the study site burned in the Hayman Fire. This fire Chenopodium spp., Fragaria spp., Rosa spp., and Solidago spp. were was ignited on 8 June, and ultimately burned more than 52,000 ha often identified only to genus, and so for consistency we included (Graham 2003). Heavy and continuous fuels, low relative humidity, them solely at this level. A small proportion of taxa could not be and strong and gusty winds enabled the fire to burn approximately identified to at least the genus level, and were disregarded herein. 24,000 ha on 9 June, largely as a contiguous high‐severity fire. While Nomenclature follows The Database (USDA NRCS, 2018). historical fires sometimes created high‐severity patches, the patches Measurements were conducted following the Hayman Fire created on this day appear to be of a size that is unprecedented over in 2003, 2004, 2005, 2006, 2007, and 2012 (1, 2, 3, 4, 5, and at least the last two to four centuries (Fornwalt et al., 2016). Less 10 years post‐fire), using the same methodologies described above extreme weather arrived on 10 June, enabling the fire to burn more (Abella & Fornwalt, 2015; Fornwalt & Kaufmann, 2014; Fornwalt heterogeneously until it was contained on 2 July. Our study site is et al., 2010). Furthermore, direct fire effects on the overstorey situated in the transitional zone between these two fire behavior and forest floor were assessed in 2003 in order to assign each plot patterns. a fire severity. Plots where ≤50% of the pre‐fire live overstorey trees were killed and where tree crown and organic surface mate‐ rial consumption were generally slight were categorized as burning 2.2 | Data collection and analyses with low severity (i.e., burned by light surface fire; ten plots). This Vascular understory plant composition at the study site was first as‐ 50% threshold was used to accommodate mortality of the smallest sessed in 1996/1997, 5–6 years before the Hayman Fire, in 20 upland size class (0–10 cm DBH); mortality of all trees >10 cm DBH in the 0.1‐ha plots, where uplands were dominated by dry conifer forest low‐severity plots did not exceed 20%, a more typical low‐sever‐ vegetation distinct from riparian vegetation plots that were sam‐ ity threshold, and accordingly basal area mortality in the low‐se‐ pled but not analyzed in this study (Fornwalt, Kaufmann, Huckaby, & verity plots was very low at <12% (Fornwalt, Stevens‐Rumann, & Stohlgren, 2009; Fornwalt, Kaufmann, Huckaby, Stoker, & Stohlgren, Collins, 2018). Moderate‐severity plots experienced >50% mortal‐ 2003). Six plots were sampled in 1996 and fourteen plots were sam‐ ity of live overstorey trees and had modest tree crown and organic pled in 1997. We did not detect differences among pre‐fire sampling surface material consumption (i.e., burned by moderate to severe 4 STEVENS et al. | Journal of Vegetation Science surface fire; six plots). High‐severity plots experienced 100% live plant lineages in the fossil record (Raven & Axelrod, 1978), an ap‐ overstorey tree mortality and complete or nearly complete tree proach developed for the California flora (Harrison & Grace, 2007) crown and organic surface material consumption (i.e., burned by that we applied to our dataset. Biogeographic affinity describes the severe crown fire; four plots). general location and associated climate where distinct evolutionary The three burn severity classes and the sampling year were our lineages (often at the genus level) first appear in the fossil record primary predictor variables for our analyses of plant community (Raven & Axelrod, 1978). For the purposes of our simplified classi‐ composition (see below), but because pre‐fire environmental factors fication of Raven and Axelrod's (1978) original work, we distinguish can influence burn severity, we also tested how such factors varied the “cool‐mesic” biogeographic affinity, largely comprising genera with burn severity class. Data on overstorey tree structure, ground that diversified during the early Tertiary (ca. 56–34 Ma) at higher lat‐ cover, and topography were collected in 1996/1997 (Fornwalt et al., itudes in North America during relatively wetter conditions (“Arcto‐ 2003, 2009; Kaufmann, Regan, & Brown, 2000). Overstorey mea‐ Tertiary” flora in Raven & Axelrod, 1978), from the “warm‐xeric” surements included species, diameter at breast height, total height, biogeographic affinity, largely comprising genera that diversified and live or dead status for all trees >1.4 m tall. From these data, we during the later Tertiary (ca 23–1.8 Ma) at lower latitudes in North calculated live stand density (trees per ha) and live basal area (m2 per America, particularly in the southwestern US and Mexico, during rel‐ ha). We also used these data and the Central Rockies Variant of the atively drier conditions (Stevens et al., 2015). Our warm‐xeric group Forest Vegetation Simulator (FVS) to estimate percent canopy cover includes three distinct groups classified by Raven and Axelrod (1978) (Dixon, 2002, p. 226; Keyser & Dixon, 2008). Ground cover mea‐ as "Madro‐Tertiary" (semiarid climate), "California Floristic Province" surements included percent bare soil cover, litter and duff cover, and (mediterranean climate), and "Warm Temperate Desert" (arid cli‐ wood cover, which were recorded for 10 systematically located 1‐m2 mate) in origin (Harrison & Grace, 2007). subplots per plot and averaged. Topography measurements included Our dataset contained 188 unique taxa, of which 171 were iden‐ elevation, slope and aspect. Aspect was categorized for analysis as tified to the species level and 17 to the genus level. Of these 188 southwest if it was between 135° and 315°, reflecting a more xeric unique taxa, 166 were native to the continental USA, and of the 166 exposure, otherwise it was categorized as northeast. We tested for native taxa, we were able assign 154 (93%) to a biogeographic affin‐ differences among the three eventual burn severity classes for the ity. Of the 154 classified taxa, 63 (41%) were species that had exact following pre‐fire environmental variables: elevation, aspect, slope, matches to species in previous California‐based datasets where bio‐ percent bare soil cover, litter and duff cover and coarse woody de‐ geographic affinities had already been classified (Harrison & Grace, bris cover, live tree density, live tree basal area, and live tree canopy 2007; Raven & Axelrod, 1978; Stevens et al., 2015), and 64 (42%) cover. We used pairwise t tests to compare the classes for all vari‐ belonged to genera in these datasets that Raven and Axelrod (1978) ables except aspect, where we used a χ2 test to determine whether classified at the genus level (we assumed the same genus‐level clas‐ the distribution of a categorical aspect variable differed among burn sification held in Colorado). We classified the remaining 27 taxa severity classes. (17%) using the biogeographic affinities of close relatives based on To assess understory plant community composition over time existing phylogenies and associations with plant families that had a and across the fire severity gradient, we categorized all understory single classification in the Raven and Axelrod (1978) scheme, or by taxa by their biogeographic affinity. The biogeographic affinity clas‐ identifying taxonomic name changes. We could not confidently clas‐ sifications describe the time period and geographic origin of distinct sify 12 (7%) of the 166 native taxa.

Fire severity Low Moderate High

40

30

chnes s Biogeographic affinity Cool−mesic 20 Warm−xeric Mean plot ri FIGURE 2 Mean plot‐level richness (number of taxa per 0.1‐ha plot) by 10 biogeographic affinity and fire severity. Error bars represent one standard error around the plot‐level mean. Dashed 2000 2005 2010 2000 20052010 2000 2005 2010 vertical line indicates the year of the Year Hayman Fire STEVENS et al. 5 Journal of Vegetation Science |

We assessed trends over time in the richness (per 0.1‐ha plot) and the same plot. The year of the permanent extinction was defined proportion of understory taxa belonging to the cool‐mesic versus as the first year a taxon was absent in a plot between 2003 (the warm‐xeric floristic groups, as well as the effects of fire severity on first post‐fire census) and 2007 (the fifth post‐fire census), inclusive. these trends. We used linear mixed‐effects models that assigned a We defined a permanent colonization similarly, where a taxon was random intercept for plot, thereby accounting for repeated sampling absent from a given plot in 1997 and present in 2012. The year of of the same plots over time by allowing a given plot to have higher or permanent colonization was defined as the first year a taxon was lower overall values of the response variables, using the R package present in a plot between 2003 and 2007, inclusive. We recognize lme4 (Bates, Maechler, Bolker, & Walker, 2013). We evaluated the that we use the term “permanent” to only refer to changes evident in significance of these trends using the Kenward–Rogers approxima‐ the last sampling year (2012), and that subsequent changes in these tion to estimate degrees of freedom in the mixed‐effects models, individual taxa may have occurred. and then compared the t statistic for the relevant fixed‐effect coef‐ ficient in the model against a t distribution for the given degrees of freedom via the R package pbkrtest (Halekoh & Højsgaard, 2014). 3 | RESULTS We also estimated extinctions and colonizations of each unique taxon at the plot level over the 15‐year duration of this study. We Richness of the cool‐mesic floristic group was consistently higher were particularly interested in qualitatively comparing extinctions than richness of the warm‐xeric floristic group, across all years and colonizations that were “permanent” for the duration of the (before and after the Hayman Fire) and all fire severities (Figure 2; study (i.e., persisting at least 10 years post‐fire). We defined a per‐ t = 38.7, df = 109, p < 0.001). Within the cool‐mesic floristic group, manent extinction as a taxon that was present in the pre‐fire census richness increased in the first year after low‐ and moderate‐sever‐ in 1997, and was absent from the final post‐fire census in 2012 from ity burns and stayed higher than pre‐fire levels up to 10 years after fire, with a plateau in richness occurring between roughly four and 10 years post‐fire (Figure 2). However, the north‐temperate floristic (a) 0.9 group experienced a decrease in richness in the first year after high‐ severity fire, which persisted for two years before increasing rapidly in year three and reaching a similar plateau as in the low‐ and moder‐ 0.8 Fire severity ate‐severity plots. The increase in cool‐mesic richness over time was

Low significant for low‐severity plots (t = 5.9, df = 59, p < 0.001) and for

ion of flora Moderate moderate‐severity plots (t = 5.7, df = 35, p < 0.001), and marginally 0.7 High significant for high‐severity plots (t = 1.8, df = 23, p = 0.08), which

Propo rt had high variance due to a small sample size but nonetheless was witn cool−mesic affinity robust to the removal of one plot with a very low ratio of cool‐mesic 0.6 to warm‐xeric taxa (t = 2.0, df = 17, p = 0.06 without the outlier). The warm‐xeric floristic group, conversely, saw an immediate increase 2000 2004 2008 2012 in richness one year post‐fire for all three fire severities (Figure 2; Year (b) for low, moderate and high severity, respectively: t = 3.2, 5.6, 6.0; df = 59, 35, 23; all p < 0.001). 0.85 Consistent with the strong temporal increase in the richness of warm‐xeric taxa in high‐severity stands and the lack of significant 0.80 Fire severity temporal increase in cool‐mesic taxa in high‐severity stands, we ob‐

ion of flora Low served greater thermophilization (a decrease in the ratio of cool‐mesic rt 0.75 Moderate to warm‐xeric taxa) over time in stands that burned at higher severity High (Figure 3a). Specifically, the relative proportion of cool‐mesic taxa de‐

0.70 creased strongly and significantly in high‐severity plots over the 15‐ witn cool−mesic affinity Mean propo year duration of this study (t = 5.3, df = 23, p < 0.001; t = 3.9, df = 17, p = 0.001 without the outlier identified above). The decrease in the 0.65 proportion of cool‐mesic taxa in the moderate severity plots was also 2000 2005 2010 Year significant (t = 3.47, df = 35, p = 0.001), while there was no significant trend in the low‐severity plots (Figure 3a; t = 0.79, df = 59, p = 0.43). FIGURE 3 Change over time in proportion of total plot richness The high‐severity and moderate‐severity plots had similar decreases (number of taxa per 0.1‐ha plot) in the cool‐mesic group (vs warm‐ in the proportion of cool‐mesic taxa in the year immediately follow‐ xeric group, the two groups summing to 1) by fire severity class. ing the fire, but the decrease in this proportion continued in high‐se‐ We show linear trend (a) and annual mean trend at the plot level (b). Error bars in (b) represent one standard error around the plot‐level verity plots while it stabilized in moderate‐severity plots (Figure 3b), mean. Dashed vertical line indicates the year of the Hayman Fire despite both severity classes having nearly identical proportions of 6 STEVENS et al. | Journal of Vegetation Science

Cool−mesic Warm−xeric

2

0 No. of colonizations

−2

Fire severity −4 Low Moderate

No. of extinctions High −6 2002 2003 2004 2005 2006 2007 2002 2003 2004 2005 2006 2007 Year

FIGURE 4 Mean number of permanent extinctions (negative bars) and colonizations (positive bars) per 0.1‐ha plot over the five years post‐fire, for cool‐mesic taxa (left panel) and warm‐xeric taxa (right panel). Error bars represent one standard error around the plot‐level mean. Permanent extinctions are counted as taxa that were present in the pre‐fire inventory at a given plot, were not present for the first time in the inventory of a given post‐fire year, and did not re‐appear by the final census in 2012. Permanent colonizations are counted as taxa that were absent in the pre‐fire inventory at a given plot, appeared for the first time in a given post‐fire year, and remained present in the plot during the final census in 2012. Dashed vertical line indicates the year of the Hayman Fire

cool‐mesic taxa prior to the fire. The lack of a significant linear trend differences in pre‐fire forest structure may have been related to in the low‐severity plots was due to the recovery in the proportion of differences in eventual fire severity and in pre‐fire plant commu‐ cool‐mesic taxa in these plots to pre‐fire levels within 10 years of the nity composition. In particular, plots that eventually burned at low fire (Figure 3b), although the proportion of cool‐mesic taxa started at severity had a tendency to have increased bare ground cover and a lower level in plots that burned at low severity. reduced litter cover, tree density, basal area and canopy cover com‐ The substantial decrease in the proportion of cool‐mesic taxa pared to plots that eventually burned at moderate or high severity, in high‐severity plots over time (Figure 3) was driven by a contin‐ although these differences were rarely significant (Appendix S2) and ued increase in the warm‐xeric richness, which by 2012 was higher differences in weather also likely played a strong role in eventual than in the low‐severity plots despite having lower richness than fire severity (Fornwalt et al., 2018). However, notably, the consider‐ the low‐severity plots prior to the fire (Figure 2). In general, it was able variation in the pre‐fire ratio of cool‐mesic to warm‐xeric taxa more common for novel warm‐xeric taxa to colonize after the fire (Figure 3a) is well explained by variation in pre‐fire canopy cover, than for previously present warm‐xeric taxa to go extinct (Figure 4; with higher canopy cover plots associated with significantly higher Table 1), with this colonization process more pronounced in moder‐ ratios of cool‐mesic to warm‐xeric taxa (Figure 5). ate and high‐severity plots than in low‐severity plots. While “perma‐ nent extinctions” (i.e., taxa present pre‐fire failing to recover within 10 years of the fire in these stands; Table 2) were uncommon among 4 | DISCUSSION warm‐xeric taxa, they were much more common in cool‐mesic taxa, particularly in the first year after the fire (Figure 4). Permanent Increasingly severe fire effects in Colorado's Hayman Fire were as‐ extinctions of cool‐mesic taxa were also much more common in sociated with increasingly pronounced shifts in understory plant high‐severity plots than in low‐severity plots, with moderate‐se‐ composition in comparison to pre‐fire communities (Figures 3,4). verity plots having an intermediate level of permanent extinctions Specifically, we found strong evidence of greater thermophilization (Figure 4), although it is possible that some of the extinctions oc‐ over time with greater fire severity, indicated by shifts in the com‐ curred before the fire. munity from taxa with cool‐mesic biogeographic affinities towards Evidence from plot environmental data suggests that the tempo‐ taxa with warm‐xeric biogeographic affinities. While shifts towards ral trend toward increased thermophilization at higher severities was warm‐xeric understory taxa with increasing disturbance sever‐ largely driven by changes in forest structure. The plots in the three ity have been documented previously within three years of fire in severity classes did not differ significantly from each other in ele‐ California (Stevens et al., 2015), we show for the first time that, in vation, aspect or slope (Appendix S2), and yet exhibited strong dif‐ Colorado, colonizing warm‐xeric taxa are largely observed within the ferences in post‐fire floristic pathways (e.g., Figure 3). Furthermore, first year after fire, representing a marked departure from pre‐fire STEVENS et al. 7 Journal of Vegetation Science |

TABLE 1 List of taxa permanently Number of plots colonized, by colonizing during the first year post‐fire severity class (subsequent colonizations not shown), with cool‐mesic (n = 22) and warm‐xeric Scientific name High Moderate Low Biogeographic affinity (n=16) biogeographic affinities Allium cernuum 0 1 2 Cool‐mesic Androsace septentrionalis 1 2 1 Cool‐mesic Antennaria parvifolia 0 0 1 Cool‐mesic Arabis fendleri 1 0 2 Cool‐mesic Artemisia ludoviciana 0 0 1 Cool‐mesic Campanula rotundifolia 1 0 1 Cool‐mesic Elymus elymoides 0 0 2 Cool‐mesic Erigeron subtrinervis 0 2 0 Cool‐mesic Erysimum capitatum 0 1 2 Cool‐mesic Heterotheca villosa 0 1 0 Cool‐mesic Packera fendleri 1 0 3 Cool‐mesic Penstemon glaber 0 2 3 Cool‐mesic Penstemon virens 0 1 0 Cool‐mesic Poa fendleriana 0 1 0 Cool‐mesic Potentilla hippiana 0 1 1 Cool‐mesic Ribes cereum 0 0 1 Cool‐mesic deliciosus 0 0 1 Cool‐mesic Scutellaria brittonii 0 0 1 Cool‐mesic Silene scouleri 0 0 1 Cool‐mesic Solidago 1 0 2 Cool‐mesic Symphoricarpos albus 0 0 1 Cool‐mesic Townsendia grandiflora 0 0 2 Cool‐mesic Aliciella pinnatifida 0 0 1 Warm‐xeric Bahia dissecta 0 3 1 Warm‐xeric Cercocarpus montanus 0 0 1 Warm‐xeric Chamerion angustifolium 0 1 0 Warm‐xeric Chenopodium 4 3 3 Warm‐xeric Conyza canadensis 0 1 0 Warm‐xeric Cryptantha virgata 1 0 1 Warm‐xeric Eriogonum alatum 1 2 0 Warm‐xeric Ipomopsis aggregata 0 1 1 Warm‐xeric Machaeranthera bigelovii 0 2 1 Warm‐xeric Pediocactus simpsonii 0 0 1 Warm‐xeric Phacelia heterophylla 2 4 3 Warm‐xeric Yucca glauca 1 0 0 Warm‐xeric Pediocactus simpsonii 0 1 0 Warm‐xeric Verbena 0 0 1 Warm‐xeric

Note: Number of plots is indicated by severity class, out of a possible total of ten low‐severity plots, six moderate‐severity plots, and four high‐severity plots. plant community composition. We also show that these taxa can functional tolerance for water stress (Harrison & Grace, 2007; persist for at least 10 years post‐fire. Stevens et al., 2015), corresponding to the increasing canopy open‐ The plant community shifts we observed are likely associated ness, understory temperatures, wind speeds, and evapotranspira‐ with changes in the microclimate with increasing disturbance se‐ tional demand increase created by canopy tree mortality (Ma et al., verity (North, Oakley, Fiegener, Gray, & Barbour, 2005). The warm‐ 2010). Thus, the microclimatic shifts associated with high‐severity xeric biogeographic affinity is generally associated with a greater fire are likely filtering the regional species pool to favor an understory 8 STEVENS et al. | Journal of Vegetation Science

TABLE 2 List of taxa going Number of plots extinct, by severity permanently extinct during the first class year post‐fire (subsequent extinctions Scientific name High Moderate Low Biogeographic affinity not shown), with cool‐mesic (n = 29) and warm‐xeric (n=9) biogeographic affinities Achnatherum scribneri 1 0 0 Cool‐mesic Arenaria hookeri 0 1 0 Cool‐mesic Campanula rotundifolia 1 0 0 Cool‐mesic Castilleja integra 1 0 0 Cool‐mesic Clematis columbiana 1 0 0 Cool‐mesic Crepis 1 0 0 Cool‐mesic Elymus trachycaulus 0 1 1 Cool‐mesic Erigeron compositus 1 0 0 Cool‐mesic saximontana 0 1 0 Cool‐mesic Fragaria 1 0 1 Cool‐mesic Gentiana 0 0 1 Cool‐mesic Hesperostipa comata 0 0 1 Cool‐mesic Heuchera parvifolia 1 0 0 Cool‐mesic Hieracium fendleri 1 0 0 Cool‐mesic Juniperus communis 3 3 4 Cool‐mesic Leucocrinum montanum 0 1 0 Cool‐mesic Leucopoa kingii 1 1 0 Cool‐mesic Oxytropis 0 0 1 Cool‐mesic Poa 1 0 0 Cool‐mesic Poa annua 0 1 0 Cool‐mesic Poa reflexa 1 2 0 Cool‐mesic Potentilla hippiana 1 0 1 Cool‐mesic Ribes inerme 0 0 1 Cool‐mesic Rosa 0 1 1 Cool‐mesic Rubus deliciosus 1 1 0 Cool‐mesic Sedum lanceolatum 1 0 0 Cool‐mesic Senecio 0 0 1 Cool‐mesic Solidago 0 0 1 Cool‐mesic Symphoricarpos albus 0 1 0 Cool‐mesic Astragalus parryi 0 1 0 Warm‐xeric Blepharoneuron tricholepis 1 0 0 Warm‐xeric Bouteloua gracilis 0 0 1 Warm‐xeric Collomia linearis 0 0 1 Warm‐xeric Cryptantha virgata 1 0 0 Warm‐xeric Gayophytum diffusum 0 0 1 Warm‐xeric Oenothera caespitosa 0 1 0 Warm‐xeric Pediocactus simpsonii 0 1 0 Warm‐xeric Verbena 0 0 1 Warm‐xeric

Note: Number of plots is indicated by severity class, out of a possible total of ten low‐severity plots, six moderate‐severity plots, and four high‐severity plots. flora associated with warmer, drier conditions. A component of taxa returned to pre‐burn levels within 10 years, while it remained this warm‐xeric flora may also be adapted to surface disturbance, low in moderate‐ and high‐severity stands relative to pre‐fire levels, and indeed we saw a shift towards the warm‐xeric flora following and comparable to low‐severity stands in absolute terms (Figure 3). low‐severity burning where the canopy was minimally affected. In The strong and persistent thermophilization observed in stands that low‐severity stands, however, the ratio of cool‐mesic to warm‐xeric experienced moderate‐ and especially high‐severity fire suggests STEVENS et al. 9 Journal of Vegetation Science |

FIGURE 5 Effects of pre‐fire canopy cover on the proportion of pre‐fire G

understory flora in the cool‐mesic y 0.9 G category (vs. the warm‐xeric category). G G G Relative pre‐fire richness of cool‐mesic flora was significantly and positively G Fire severity G Low related to pre‐fire canopy cover, both over G G G Moderate G G 0.8 G all plots and over eventual low‐severity G G High G plots only tion of pre−fire flora G G G witn cool−mesic affinit

Propor G G 0.7 G G

25 30 35 40 45 50 Pre−fire canopy cover (%) that canopy removal by fire contributed to the strong differences in Lutz, 2019; Rapacciuolo et al., 2014). Supporting this inference, we post‐fire communities among severity classes. Recent research also showed that extinctions of cool‐mesic lineages were most pronounced suggests that altered microclimates caused by high‐severity fire in in high‐severity burned areas (Figure 4), which had complete canopy dry forests may lead to long‐term diversity losses of lichens, another removal at the plot scale. Although our post‐fire timeframe of 10 years taxonomic group that is characteristic of cooler, more mesic environ‐ may limit our ability to infer longer‐term community dynamics, we be‐ ments (Miller, Root, & Safford, 2018). lieve that the changes we observed in canopy structure and under‐ Interestingly, the ratio of cool‐mesic to warm‐xeric taxa in 1997, story diversity should persist for the near future due in large part to prior to the 2002 Hayman Fire, was actually higher in stands that the slow growth rates of surviving and establishing trees (Fornwalt subsequently burned at moderate and high severity compared to et al., 2018; Malone et al., 2018). low‐severity stands (Figure 3). While there was no clear evidence that The strength of our conclusions is somewhat limited by the rela‐ pre‐fire stand density or basal area were greater in stands that even‐ tively small sample size of plots due to the opportunistic nature of our tually burned at moderate or high severity compared to low severity study, particularly in high‐severity stands which predictably had high (Appendix S2), we found that higher pre‐fire canopy cover was strongly variance in understory data (e.g., Figure 2). However, despite this small associated with an understory dominated by cool‐mesic taxa (Figure 5), sample size we still detected significantly greater thermophilization in consistent with previous research suggesting that variation in microcli‐ high‐severity burn classes compared with low‐ and moderate‐severity mates created by forest structure is largely responsible for understory classes (Figures 2,3). Our analyses of temporal trends in understory thermophilization (De Frenne et al., 2013; Stevens et al., 2015). vegetation are strengthened by the inclusion of pre‐fire understory This study provides insight into the relative stability of plant com‐ data, which are uncommon (Fornwalt & Kaufmann, 2014), and multiple munities in dry conifer forests in the first decade following low‐sever‐ lines of evidence support the importance of the overstorey canopy ity burning. Communities in low‐severity burns appear more resistant in driving understory thermophilization (e.g., Figure 5). Alternatively, to disturbance than those in high‐severity burns over time, in that they the thermophilization observed over time could be attributable to exhibit less change in composition, and more resilient, in that they generally below‐average precipitation during the post‐fire years in this move toward their pre‐disturbance composition relatively quickly, region (Appendix S1), but we did not observe this trend in low‐severity at least with respect to the ratio of the two biogeographic groups plots despite their experiencing the same drying trend (Figure 3), again (Figure 3). These trends in understory thermophilization are consis‐ suggesting that canopy loss is the primary driver of change. It is also tent with our expectation that low‐ and moderate‐severity burns (at possible that plots that eventually burned at high severity had a greater the sub‐hectare scale) are associated with greater structural resilience abundance of warm‐xeric taxa in the surrounding landscape and this in dry forests (Stevens, Safford, & Latimer, 2014) where understory was the main driver of thermophilization. We think this is less likely structure and tree regeneration are dependent on a local component than a canopy‐driven explanation, because these high‐severity plots of surviving trees. Critically, the heterogeneity in understory microcli‐ had the highest pre‐fire proportion of cool‐mesic taxa (Figures 2,3) mate and species composition is facilitated by the fine‐scale (i.e., sub‐ and generally higher pre‐fire canopy cover Figure 5). hectare, individual tree scale) heterogeneity of canopy cover occurring The plant biogeographic affinity classifications pioneered by Raven within stands with low‐ and moderate‐severity disturbance at that and Axelrod (1978) has proven to be a useful framework for understand‐ scale (Dodson & Peterson, 2010; Stevens et al., 2015). Post‐fire can‐ ing California plant communities, and this study suggests the frame‐ opy heterogeneity within low‐ and moderate‐severity stands (which work may also be applicable to other locales in western North America. retained some canopy cover in our study, unlike high‐severity plots) Although Raven and Axelrod's (1978) floristic classifications rely on the appears to be important for the greater persistence of cool‐mesic historical concept of geofloras, which pre‐date contemporary phyloge‐ lineages in post‐fire stands, via the creation of microclimatic refugia netic methods, numerous quantitative studies lend credence to their within warm and dry conifer forests (Blomdahl, Kolden, Meddens, & classifications. For example, woody cool‐mesic plants have been shown 10 STEVENS et al. | Journal of Vegetation Science to have significantly larger seeds and lower specific area than woody REFERENCES warm‐xeric plants (Ackerly, 2003). Raven and Axelrod's classifications Abella, S. R., & Fornwalt, P. J. (2015). Ten years of vegetation assembly have also been shown to reflect contemporary plant distributions after a North American mega fire. Global Change Biology, 21, 789– among ecoregions in California (Ackerly, 2009; Harrison & Grace, 2007), 802. https​://doi.org/10.1111/gcb.12722​ and there is evidence that cool‐mesic species have declined more over Ackerly, D. D. (2003). Community assembly, niche conservatism, and recent decades under climate change (Damschen, Harrison, & Grace, adaptive evolution in changing environments. International Journal of Plant Sciences, 164, S165–S184. https​://doi.org/10.1086/368401 2010). In this study, the extinctions of cool‐mesic taxa after fire occur Ackerly, D. D. (2009). Evolution, origin and age of lineages in the across numerous disparate phylogenetic lineages (e.g., , Lamiales, Californian and Mediterranean floras. Journal of Biogeography, 36, , and Asterales; Table 2), highlighting that patterns we observed 1221–1233. https​://doi.org/10.1111/j.1365-2699.2009.02097.x do not seem to be caused by the chance vulnerabilities of a few taxa to Asherin, L. A. (2016). Manitou experimental forest hourly meteorology data, 2nd ed. Fort Collins, CO:Forest Service Research Data Archive. fire. Future research should further elucidate mechanisms linking plant Updated 05 February 2019. https​://doi.org/10.2737/rds-2011-0001-2 community shifts to fire severity through studies of functional traits and Bates, D. M., Maechler, M., Bolker, B., & Walker, S. (2013). lme4: Linear quantitative community phylogenetic patterns. mixed‐effects models using Eigen and S4. R package version 1.0‐5. Our findings suggest that management activities that reduce fire CRAN.R‐project.org/package=lme4 Battaglia, M. A., Gannon, B., Brown, P. M., Fornwalt, P. J., Cheng, A. S., & severity could contribute to the conservation of plant diversity in dry Huckaby, L. S. (2018). Changes in forest structure since 1860 in pon‐ conifer forests. Many such forests are increasingly experiencing un‐ derosa pine dominated forests in the Colorado and Front characteristically extensive high‐severity fire (Singleton et al., 2019; Range, USA. Forest Ecology and Management, 422, 147–160. https​:// Stevens et al., 2017), due at least in part to increased forest density doi.org/10.1016/j.foreco.2018.04.010 Blomdahl, E. M., Kolden, C. A., Meddens, A. J. H., & Lutz, J. A. (2019). and homogeneity following a century of fire suppression, livestock The importance of small fire refugia in the central Sierra Nevada, grazing, and logging (Battaglia et al., 2018; Safford & Stevens, 2017). California, USA. Forest Ecology and Management, 432, 1041–1052. Restoration and fuels reduction treatments that create more open https​://doi.org/10.1016/j.foreco.2018.10.038 and heterogeneous forest conditions have been shown to reduce Brown, P. M., Kaufmann, M. R., & Shepperd, W. D. (1999). Long- term, fire‐caused tree mortality, even under extreme weather conditions landscape patterns of past fire events in a montane ponderosa pine forest of central Colorado. Landscape Ecology, 14, 513–532. (Fulé, Crouse, Roccaforte, & Kalies, 2012; Safford, Stevens, Merriam, Coop, J. D., Massatti, R. T., & Schoettle, A. W. (2010). Subalpine veg‐ Meyer, & Latimer, 2012). By conserving living trees in the footprint etation pattern three decades after stand‐replacing fire: effects of of future fires, such heterogeneous treatments should also conserve landscape context and topography on plant community composition, cooler, more mesic microclimatic conditions, and the cool‐mesic taxa tree regeneration, and diversity. Journal of Vegetation Science, 21, 472–487. https​://doi.org/10.1111/j.1654-1103.2009.01154.x that are adapted to them (Stevens et al., 2015). Damschen, E. I., Harrison, S., & Grace, J. B. (2010). Climate change effects on an endemic‐rich edaphic flora: resurveying Robert H. Whittaker's Siskiyou sites (, USA). Ecology, 91, 3609–3619. https​://doi. ACKNOWLEDGEMENTS org/10.1890/09-1057.1 De Frenne, P., Rodríguez‐Sánchez, F., Coomes, D. A., Baeten, L., Verstraeten, The initial conception of this manuscript occurred at the Natural G., Vellend, M., … Verheyen, K. (2013). Microclimate moderates plant Areas Conference, Davis, CA, 2016. We thank Susan Harrison and responses to macroclimate warming. Proceedings of the National Brian Anacker for compiling and sharing initial classifications from Academy of Sciences of the United States of America, 110, 18561–18565. the Raven and Axelrod (1978) monograph. https​://doi.org/10.1073/pnas.13111​90110​18561-18565.10.1073/ pnas.13111​90110​ Dixon, G. E. (2002). Essential FVS: a user's guide to the Forest Vegetation AUTHOR CONTRIBUTIONS Simulator. Fort Collins, CO: USDA Forest Service, Forest Service Management Center. JTS and PJF conceived of the study. PJF provided the data. JTS ana‐ Dodson, E. K., & Peterson, D. W. (2010). Dry coniferous forest restoration and understory plant diversity: The importance of community hetero‐ lyzed the data. JTS, JEDM, and PJF wrote the manuscript. geneity and the scale of observation. Forest Ecology and Management, 260, 1702–1707. https​://doi.org/10.1016/j.foreco.2010.08.012 Fornwalt, P. J., Huckaby, L. S., Alton, S. K., Kaufmann, M. R., Brown, P. DATA ACCESSIBILITY M., & Cheng, A. S. (2016). Did the 2002 Hayman Fire, Colorado, USA, burn with uncharacteristic severity. Fire Ecology, 12, 117–132. https​ Final data and R code used to generate the manuscript are pub‐ ://doi.org/10.4996/firee​cology.1203117 licly available via the USDA Forest Service archive at https​://doi. Fornwalt, P. J., & Kaufmann, M. R. (2014). understory plant community org/10.2737/rds-2019-0031. dynamics following a large, mixed severity wildfire in a Pinus pondero‐ sa–Pseudotsuga menziesii forest, Colorado, USA. Journal of Vegetation Science, 25, 805–818. https​://doi.org/10.1111/jvs.12128​ ORCID Fornwalt, P. J., Kaufmann, M. R., Huckaby, L. S., & Stohlgren, T. J. (2009). Effects of past logging and grazing on understory plant communities Jens T. Stevens https://orcid.org/0000-0002-2234-1960 in a montane Colorado forest. Plant Ecology, 203, 99–109. https​://doi. org/10.1007/s11258-008-9513-z Jesse E. D. Miller https://orcid.org/0000-0002-2316-779X Fornwalt, P. J., Kaufmann, M. R., Huckaby, L. S., Stoker, J. M., & Paula J. Fornwalt https://orcid.org/0000-0002-5252-408X Stohlgren, T. J. (2003). Non‐native plant invasions in managed and STEVENS et al. 11 Journal of Vegetation Science |

protected ponderosa pine/Douglas‐fir forests of the Colorado Front climate change in California. Global Change Biology, 20, 2841–2855. Range. Forest Ecology and Management, 177, 515–527. https​://doi. https​://doi.org/10.1111/gcb.12638​ org/10.1016/s0378-1127(02)00456-5 Raven, P. H., & Axelrod, D. I. (1978). Origin and relationships of the Fornwalt, P. J., Kaufmann, M. R., & Stohlgren, T. J. (2010). Impacts of California flora. Berkeley, CA: University of California Press. mixed severity wildfire on exotic plants in a Colorado ponderosa Romme, W. H., Veblen, T. T., Kaufmann, M. R., Sherriff, R., & Regan, C. pine–Douglas‐fir forest. Biological Invasions, 12, 2683–2695. https​:// M. (2003). Ecological effects of the Hayman Fire ‐ Part 1: Historical doi.org/10.1007/s10530-009-9674-2 (pre‐1860) and current (1860‐2002) fire regimes. In R. T. Graham Fornwalt, P. J., Stevens‐Rumann, C. S., & Collins, B. J. (2018). Overstory struc‐ (Ed.), Hayman fire case study. USDA Forest Service General Technical ture and surface cover dynamics in the decade following the Hayman Report RMRS‐GTR‐114 (pp. 181–195). Ogden, UT: U.S. Department Fire, Colorado. Forests, 9, 152. https​://doi.org/10.3390/f9030152 of Agriculture, Forest Service, Rocky Mountain Research Station. Fulé, P. Z., Crouse, J. E., Roccaforte, J. P., & Kalies, E. L. (2012). Do thinning Romme, W. H., Whitby, T. G., Tinker, D. B., & Turner, M. G. (2016). and/or burning treatments in western USA ponderosa or Jeffrey pine‐ Deterministic and stochastic processes lead to divergence in plant dominated forests help restore natural fire behavior? Forest Ecology and communities 25 years after the 1988 Yellowstone fires. Ecological Management, 269, 68–81. https​://doi.org/10.1016/j.foreco.2011.12.025 Monographs, 86, 327–351. https​://doi.org/10.1002/ecm.1220 Gottfried, M., Pauli, H., Futschik, A., Akhalkatsi, M., Barančok, P., Benito Safford, H. D., & Stevens, J. T. (2017). Natural range of variation for yel‐ Alonso, J. L., … Grabherr, G. (2012). Continent‐wide response of low pine and mixed‐conifer forests in the Sierra Nevada, southern mountain vegetation to climate change. Nature Climate Change, 2, Cascades, and Modoc and Inyo National Forests, California, USA. In 111–115. https​://doi.org/10.1038/nclim​ate1329 General Technical Report PSW‐GTR‐256. Albany, CA: U.S. Department Graham, R.T. (2003). Hayman fire case study. [General Technical Report of Agriculture, Forest Service, Pacific Southwest Research Station. RMRS-GTR-114]. Fort Collins, CO, US: USDA Forest Service, Rocky Safford, H. D., Stevens, J. T., Merriam, K., Meyer, M. D., & Latimer, A. Mountain Research Station. M. (2012). Fuel treatment effectiveness in California yellow pine and Grime, J. P. (1977). Evidence for existence of three primary strategies in mixed conifer forests. Forest Ecology and Management, 274, 17–28. plants and its relevance to ecological and evolutionary theory. American https​://doi.org/10.1016/j.foreco.2012.02.013 Naturalist, 111, 1169–1194. https​://doi.org/10.1086/283244 Singleton, M. P., Thode, A. E., Sánchez Meador, A. J., & Iniguez, J. M. Halekoh, U., & Højsgaard, S. (2014). A Kenward‐Roger approximation and (2019). Increasing trends in high‐severity fire in the southwestern parametric bootstrap methods for tests in linear mixed models — The USA from 1984 to 2015. Forest Ecology and Management, 433, 709– R package pbkrtest. Journal of Statistical Software, 59, 1–32. 719. https​://doi.org/10.1016/j.foreco.2018.11.039 Harrison, S., & Grace, J. B. (2007). Biogeographic affinity helps explain Stevens, J. T., Collins, B. M., Miller, J. D., North, M. P., & Stephens, S. L. productivity–richness relationships at regional and local scales. (2017). Changing spatial patterns of stand‐replacing fire in California American Naturalist, 170, S5–S15. https​://doi.org/10.1086/519010 conifer forests. Forest Ecology and Management, 406, 28–36. https​:// Hart, S. A., & Chen, H. Y. H. (2006). Understory vegetation dynamics of doi.org/10.1016/j.foreco.2017.08.051 North American boreal forests. Critical Reviews in Plant Sciences, 25, Stevens, J. T., Safford, H. D., Harrison, S., & Latimer, A. M. (2015). 381–397. https​://doi.org/10.1080/00735​26008​19286​ Forest disturbance accelerates thermophilization of understory Johnstone, J. F., Allen, C. D., Franklin, J. F., Frelich, L. E., Harvey, B. J., plant communities. Journal of Ecology, 103, 1253–1263. https​://doi. Higuera, P. E., … Turner, M. G. (2016). Changing disturbance regimes, org/10.1111/1365-2745.12426​ ecological memory, and forest resilience. Frontiers in Ecology and the Stevens, J. T., Safford, H. D., & Latimer, A. M. (2014). Wildfire‐contingent Environment, 14, 369–378. https​://doi.org/10.1002/fee.1311 effects of fuel treatments can promote ecological resilience in sea‐ Kaufmann, M. R., Regan, C. M., & Brown, P. M. (2000). Heterogeneity sonally dry conifer forests. Canadian Journal of Forest Research, 44, in ponderosa pine/Douglas‐fir forests: age and size structure in un‐ 843–854. https​://doi.org/10.1139/cjfr-2013-0460 logged and logged landscapes of central Colorado. Canadian Journal USDA NRCS. (2018). The PLANTS database. Baton Rouge, LA, US: of Forest Research, 30, 698–711. https​://doi.org/10.1139/x99-255 National Plant Data Center. http://www.plants.usda.gov Keyser, C. E., & Dixon, G. E. (2008). Central Rockies (CR) variant overview Webster, K. M., & Halpern, C. B. (2010). Long‐term vegetation responses ‐ Forest Vegetation Simulator. Ft Collins, CO: USDA Forest Service, to reintroduction and repeated use of fire in mixed‐conifer forests Forest Management Service Center. of the Sierra Nevada. Ecosphere, 1, 1–27. https​://doi.org/10.1890/ Ma, S., Concilio, A., Oakley, B., North, M., & Chen, J. (2010). Spatial vari‐ es10-00018.1 ability in microclimate in a mixed‐conifer forest before and after thin‐ ning and burning treatments. Forest Ecology and Management, 259, 904–915. https​://doi.org/10.1016/j.foreco.2009.11.030 SUPPORTING INFORMATION Malone, S. L., Fornwalt, P. J., Battaglia, M. A., Chambers, M. E., Iniguez, J. M., & Sieg, C. H. (2018). Mixed‐severity fire fosters heteroge‐ Additional supporting information may be found online in the neous spatial patterns of conifer regeneration in a dry conifer forest. Supporting Information section at the end of the article. Forests, 9, 45. https​://doi.org/10.3390/f9010045 Miller, J. E. D., Root, H. T., & Safford, H. D. (2018). Altered fire regimes Variation in annual site precipitation from 1995‐2012 cause long‐term lichen diversity losses. Global Change Biology, 24, APPENDIX S1 4909–4918. https​://doi.org/10.1111/gcb.14393​ APPENDIX S2 Comparison of environmental conditions across Miller, J. E. D., & Safford, H. D. (2019). Plant community responses to plots of different burn severity wildfire may be contingent upon historical disturbance regimes. h t t p s ://doi.org/10.32942/​ osf.io/8sxfh​ ​ North, M., Oakley, B., Fiegener, R., Gray, A., & Barbour, M. (2005). How to cite this article: Stevens JT, Miller JED, Fornwalt PJ. Influence of light and soil moisture on Sierran mixed‐conifer un‐ derstory communities. Plant Ecology, 177, 13–24. https​://doi. Fire severity and changing composition of forest understory org/10.1007/s11258-005-2270-3 plant communities. J Veg Sci. 2019;00:1–11. h t t p s :​ / / d o i . Rapacciuolo, G., Maher, S. P., Schneider, A. C., Hammond, T. T., Jabis, M. org/10.1111/jvs.12796​ D., Walsh, R. E., … Beissinger, S. R. (2014). Beyond a warming fin‐ gerprint: individualistic biogeographic responses to heterogeneous